mcp-malt
Enables access to the documented Malt API for back-office tasks, including retrieving invoices, payments, and commission invoices, as well as SCIM user provisioning for organization accounts.
README
mcp-malt
An MCP server for the Malt API. It covers all 13 operations Malt publishes: the invoices you issue, the commission Malt bills you, the payments you receive, and SCIM user provisioning for organization accounts.
What this does and does not cover
Malt's API is smaller than most people expect. It covers back office billing and enterprise user provisioning, and nothing else:
| Toolset | Tools | What it reaches |
|---|---|---|
invoices |
3 | Invoices you issued to clients, including the PDF |
payments |
1 | Payments received, with the invoices each one settles |
fee-invoices |
3 | Malt's commission invoices to you, including the PDF |
scim |
2 read, 4 write | SCIM 2.0 user provisioning for an organization account |
There is no talent search, no profiles, no missions, no offers, and no messaging. Those are what Malt is known for, and none of them are in the public API. If that is what you came for, this server cannot give it to you, and neither can anything else built on the documented API.
Two community MCP servers for Malt do reach that data, by driving the web app with browser automation and scraping the DOM. Their own documentation warns that automating profile edits may breach Malt's terms of service. This server only calls the documented API.
Getting a token
- Create an identity on the signup page if you do not have one.
- Open My Account, then API Keys.
- Create an access token with the permission scopes you need. Match them to the toolsets you plan to enable.
- Copy the token straight away. It is shown once, at creation, and never again.
Freelancer account tokens are self served. Client team and organization tokens, which the scim toolset needs, come through your Malt representative.
There is no OAuth 2.0 here. Malt publishes no authorization server and no token endpoint, so the API token above is the only way in. docs/authentication.md explains this in more detail, including the header detail that trips most people up.
Install
Published three ways from one release: the npm package on npmjs and on GitHub Packages, and the container image on GHCR. npx is the usual choice for an MCP client; Docker suits a locked-down or shared setup.
Pin a version in the examples below. An MCP server runs with your API token, so an unpinned @latest means a new release starts running on your machine without you choosing it.
npm install @ialejandro/mcp-malt@1.0.0
To move to a newer version, bump the number in your client config and restart the client. The releases page lists what changed in each one.
To install from GitHub Packages instead of npmjs, point the scope at it: @ialejandro:registry=https://npm.pkg.github.com in your .npmrc.
npx
// Claude Desktop: claude_desktop_config.json
// Cursor: .cursor/mcp.json
{
"mcpServers": {
"malt": {
"command": "npx",
"args": ["-y", "@ialejandro/mcp-malt@1.0.0"],
"env": {
"MALT_API_TOKEN": "your-token-here",
"MALT_TOOLSETS": "invoices,payments,fee-invoices"
}
}
}
}
Claude Code
claude mcp add malt \
--env MALT_API_TOKEN=your-token-here \
--env MALT_TOOLSETS=invoices,payments,fee-invoices \
-- npx -y @ialejandro/mcp-malt@1.0.0
Docker
The server speaks JSON-RPC over stdin and stdout, so -i is required and -t must be left off.
docker run -i --rm \
-e MALT_API_TOKEN=your-token-here \
-e MALT_TOOLSETS=invoices \
ghcr.io/ialejandro/mcp-malt:1.0.0
// Claude Desktop: claude_desktop_config.json
// Cursor: .cursor/mcp.json
{
"mcpServers": {
"malt": {
"command": "docker",
"args": [
"run", "-i", "--rm",
"-e", "MALT_API_TOKEN",
"-e", "MALT_TOOLSETS",
"ghcr.io/ialejandro/mcp-malt:1.0.0"
],
"env": {
"MALT_API_TOKEN": "your-token-here",
"MALT_TOOLSETS": "invoices,payments,fee-invoices"
}
}
}
}
From source
git clone https://github.com/ialejandro/mcp-malt
cd mcp-malt
npm ci && npm run build
{
"mcpServers": {
"malt": {
"command": "node",
"args": ["/absolute/path/to/mcp-malt/dist/index.js"],
"env": {
"MALT_API_TOKEN": "your-token-here",
"MALT_TOOLSETS": "invoices,payments,fee-invoices"
}
}
}
}
{
"mcpServers": {
"malt": {
"command": "docker",
"args": [
"run", "-i", "--rm",
"-e", "MALT_API_TOKEN",
"-e", "MALT_TOOLSETS",
"ghcr.io/ialejandro/mcp-malt:1.0.0"
],
"env": {
"MALT_API_TOKEN": "your-token-here",
"MALT_TOOLSETS": "invoices,payments,fee-invoices"
}
}
}
}
Configuration
| Variable | Default | What it does |
|---|---|---|
MALT_API_TOKEN |
required | Your token from My Account, API Keys |
MALT_TOOLSETS |
empty | Comma list of invoices, payments, fee-invoices, scim, or all |
MALT_ALLOW_WRITES |
false |
Exposes the four SCIM write tools. scim alone gives you reads only |
MALT_MAX_LIST_ITEMS |
200 |
Caps list results, since Malt does not paginate the billing endpoints |
MALT_RATE_LIMIT_RPS |
5 |
Outbound pacing. Our number, not Malt's, which publishes none |
MALT_TIMEOUT_MS |
30000 |
Per request timeout |
MALT_API_BASE_URL |
https://api.malt.com |
Override, mostly for testing |
MALT_AUTH_SCHEME |
raw |
raw or bearer. Leave it alone unless Malt changes |
MALT_LOG_LEVEL |
info |
debug, info, warn, error. All logging goes to stderr |
With no MALT_TOOLSETS the server starts and offers nothing. That is deliberate, not a failure: you decide what a model gets to see. Add toolsets one at a time.
An unknown toolset name stops the server with an error rather than being ignored, because a typo that silently costs you three tools is miserable to debug.
Common tasks
Reconcile a quarter. Enable invoices,payments,fee-invoices and run the malt_reconcile_revenue prompt. It pulls all three lists, matches payments to the invoices they settle, and reports gross billed, tax, Malt's commission, and cash received.
Reconcile my Malt revenue for Q1 2026.
Get an invoice PDF. Malt returns PDFs base64-encoded inside a JSON body. The tools decode them and hand back a real PDF attachment.
Download the PDF for invoice INV-123456.
Offboard someone. Enable scim with MALT_ALLOW_WRITES=true and use malt_user_lifecycle. It deactivates rather than deletes, which is the path Malt actually supports.
Offboard jane.doe@acme.com from our Malt organization.
Documentation
- Getting and managing a token
- Authentication, and why there is no OAuth
- Toolsets: what to enable and when
- Every endpoint, with requests and responses
- Errors and what they mean
- Rate limits
- Extending the server
Development
npm ci
npm run typecheck
npm test
npm run build
npm run check-spec # diff the live Malt spec against spec/malt-openapi.json
Tests mock fetch and never call the live API.
spec/malt-openapi.json is a committed snapshot of Malt's published document. A daily workflow diffs the live spec against it and opens a PR when anything moves. Malt has kept info.version at 0.0.1 throughout, so the check compares operations rather than trusting the version number.
Contributing
See CONTRIBUTING.md. Adding an endpoint or a toolset is described in docs/extending.md.
License
MIT
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.